Download presentation
Presentation is loading. Please wait.
1
Linear regression project
Machine Learning; Mon Apr 21, 2008
2
Motivation To get a feeling for the material so far, you now get a mandatory project. It should help you link the theory with practical machine learning problem solving.
3
The exercise Problem: You are given a data set with 500 predictor variables and corresponding 500 target values. Train a program to predict targets for new predictor variables where the targets are unknown.
4
The exercise Download training predictor and target values for the five datasets. Select your feature basis functions and construct your model matrix F. Obtain weight vector by solving FTFw=FTt. Download new predictors and use the trained model to predict corresponding targets as URL:
5
The exercise Download training predictor and target values for the five datasets. Select your feature basis functions and construct your model matrix F. Obtain weight vector by solving FTFw=FTt. Download new predictors and use the trained model to predict corresponding targets as It might be a good idea to consider more than one model i.e. set of basis functions, and select the best performing. Watch out for over-fitting! You have plenty of data, so use it for both training and test data! URL:
6
Handing it in If you choose to hand in this project, you can me your target values for the non-training data by 12:00 Fry May 9 and I will check them up against the true values.
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.